What is Anchor Browser?
Anchor Browser (b0) is an AI-powered browser automation tool. Unlike the basic Web Scraper which only fetches static HTML, Anchor Browser runs a full Chrome browser that can:
- Execute JavaScript and interact with dynamic websites
- Click buttons, fill forms, and navigate through pages
- Log into websites and access authenticated content
- Solve CAPTCHAs automatically
- Take screenshots of what it sees
- Handle sites that block traditional scrapers
- Content is in the page source
- No login required
- No JavaScript needed
- Speed is priority
- Free (no API key)
- Content loads via JavaScript
- Login/authentication needed
- Multi-step interactions
- CAPTCHAs present
- Screenshots needed
Getting Started
Step 1: Get an API Key
- Go to app.anchorbrowser.io/api-keys
- Create an account or sign in
- Generate a new API key
- In Publish Owl, go to Settings
- Find Anchor Browser (b0) and paste your API key
Step 2: Add to Your Workflow
- Create or edit a workflow
- In Workflow Steps, click + Add Step
- Select Anchor Browser (b0) from the provider dropdown
- Configure the settings (explained below)
Configuration Options
Target URL
Where the browser should start. You have three options:
- Full URL:
https://example.com/page- Start at a specific page - Keyword placeholder:
{"https://example.com/search?q={{keyword}}"}- Insert your keyword - Leave empty: The browser will Google search for your keyword
Max Steps
Controls how many actions the browser can take (clicks, scrolls, typing, etc.).
- Default: 150 steps
- Range: 1-500 steps
- Guidance: Simple tasks need 20-50 steps. Complex multi-page workflows may need 200+
AI Provider & Model
Anchor Browser uses AI to understand your instructions and control the browser. Choose which AI powers the automation:
| Provider | Example Model | Notes |
|---|---|---|
| OpenAI | gpt-5.2, o4-mini | Best overall performance |
| Anthropic | claude-sonnet-4-5 | Good for nuanced instructions |
gemini-2.5-flash | Fast and cost-effective | |
| Groq | llama-3.3-70b-versatile | Open source, very fast |
| xAI | grok-4-1-fast-reasoning, grok-4.20-0309-reasoning | Grok models |
| Azure | gpt-4o | Enterprise Azure OpenAI |
You can enter any model ID supported by the provider. Check each provider's docs for available models.
Capture Screenshot
Enable this to capture a final screenshot if none were taken during the task. This is a fallback - you can also request screenshots in your prompt (see below).
Writing Effective Prompts
The prompt tells Anchor Browser what to do. Write clear, step-by-step instructions as if you were telling someone how to use a website.
Basic Structure
1. Navigate to [where] 2. Find [what] 3. Click/Type/Extract [action] 4. Take a screenshot of [what you want to capture]
Example: Research a Product
Search for "{'{{keyword}}'}" on Amazon.
Find the top 5 products in the search results.
For each product, extract:
- Product name
- Price
- Star rating
- Number of reviews
Take a screenshot of the search results page. Example: Check Competitor Pricing
Go to competitor.com Navigate to their pricing page. Extract all pricing tiers, including: - Plan names - Monthly and annual prices - Key features for each tier Take a screenshot of the pricing table.
Example: Login and Extract Data
Go to app.example.com/login Log in with email SECRET_EMAIL and password SECRET_PASSWORD. Navigate to the dashboard. Find the analytics section. Extract the key metrics displayed. Take a screenshot of the dashboard.
Note: SECRET_EMAIL and SECRET_PASSWORD are placeholders for secure credentials. Configure the actual values in the Secret Values section of your workflow settings.
- Be specific about what to click ("Click the blue 'Sign Up' button")
- Mention "take a screenshot" when you want to capture something
- Use
{'{{keyword}}'}to insert your job's keyword - Break complex tasks into numbered steps
- Tell it what data to extract explicitly
How Phases Work
Publish Owl automatically breaks your prompt into "phases" - discrete steps that the browser executes in sequence. This happens behind the scenes.
Example Phase Breakdown
Your prompt: "Go to example.com, find the pricing page, extract prices, take a screenshot"
- Phase 1: Navigate to example.com homepage
- Phase 2: Find and click the pricing link
- Phase 3: Extract pricing information + capture screenshot
The browser session persists across phases, so login state, cookies, and page context are maintained throughout the entire task.
Capturing Screenshots
Screenshots are captured when you mention them in your prompt. Common phrases that trigger screenshots:
- "Take a screenshot of..."
- "Capture an image of..."
- "Screenshot the..."
Screenshots are stored in the cloud and included in the job results. They can be used by subsequent steps in your workflow or passed to your CMS.
Using Credentials Securely
If you need to log into a website, you can pass credentials securely using secret values. These are sent encrypted to Anchor Browser and never appear in your job logs.
How to Use
- In your prompt, use placeholder names like
SECRET_EMAILorSECRET_PASSWORD - In the Anchor Browser settings, click + Add Secret
- Enter the placeholder name (e.g.,
SECRET_EMAIL) and the actual value - The browser will substitute the real values when executing
Example Prompt
Go to app.example.com/login Log in with email SECRET_EMAIL and password SECRET_PASSWORD Navigate to the dashboard and extract the key metrics.
Then in your workflow's Secret Values section, add SECRET_EMAIL and SECRET_PASSWORD with your actual credentials.
SECRET_EMAIL and SECRET_PASSWORD instead of the actual values. Only you can see the real credentials in the workflow settings.Using with Other Steps
Anchor Browser works as part of your multi-step workflow. The data it extracts is passed to subsequent steps.
Typical Workflow
Step 1: Anchor Browser → Researches competitor pricing, captures screenshots → Output: Pricing data + screenshot URLs Step 2: OpenAI/Claude Writer → Receives the pricing data → Writes a comparison article Step 3: Editor (optional) → Polishes the article
The writing step automatically receives everything Anchor Browser extracted, including any screenshots captured.
Session Recordings
Every Anchor Browser session is recorded. As soon as the session starts, a link appears in the job logs. You can click this link to watch the browser in real-time as the browser navigates pages and completes tasks.
Look for: 🎬 View session recording: https://app.anchorbrowser.io/sessions/...
What You Can Do
- Watch live: See the browser navigate and interact with pages in real-time
- Debug issues: Understand exactly what the browser is doing and where it gets stuck
- Stop the session: If something goes wrong, you can stop the session from the Anchor Browser dashboard
- Review later: Recordings are saved so you can review completed sessions anytime
Limitations
- Not available for Template (pSEO) workflows: Anchor Browser is skipped in batch processing due to cost and time
- Session limit: Each session has a 30-minute maximum duration
- One job at a time: Each job creates a new browser session
- Costs money: Anchor Browser charges per session time and AI usage
- Slower than scraper: Full browser automation takes more time
Troubleshooting
"Anchor Browser API key not provided"
Go to Settings and add your Anchor Browser API key. Get one at app.anchorbrowser.io/api-keys
Task not completing or timing out
Try increasing Max Steps. Complex multi-page tasks need more steps. Also simplify your prompt - break very complex tasks into multiple simpler jobs.
Browser not finding elements
Be more specific in your prompt. Instead of "click the button", say "click the blue 'Submit' button in the form". Check the session recording to see what happened.
Login failing
Make sure you're using secret values for credentials. Check that the login flow matches your instructions (some sites have multi-step logins).
No screenshots captured
Make sure your prompt explicitly says "take a screenshot" or similar. Enable "Capture Screenshot" as a fallback option.
Best Practices
1. Start Simple
Test with a simple task first before building complex workflows. Verify it works, then add more steps.
2. Check Recordings
Always review session recordings for failed jobs to understand what went wrong.
3. Be Specific
Vague instructions lead to unpredictable results. Tell the browser exactly what to look for and what to do.
4. Use for Research, Not Production Scraping
Anchor Browser is great for gathering research that informs your content. For high-volume data extraction, consider the simpler Web Scraper if possible.
5. Secure Your Credentials
Always use secret values for passwords and sensitive data. Never hardcode credentials in prompts.
Built-in Features
Anchor Browser includes these features automatically:
- CAPTCHA Solving: Automatically handles most CAPTCHAs
- Stealth Mode: Uses a patched browser to avoid detection
- Residential Proxy: Routes through residential IPs for better access
- Popup Blocking: Automatically dismisses cookie banners and popups
- Session Recording: Every session is recorded for debugging